First Slice

Geography

This is the fist slice of our Indonesia implementation of POSEIDON. We have a good year of work ahead of us but it’s important to have something in a working state to expand from the earliest stages.

The geography of course is the easy bit (because Steve does it). Steve Saul grabbed the 30 arc-second bathymetry data from Gebco and in a csv format it is easy to import in POSEIDON.

We can change the resolution at will, so for example a 100 cells wide map would look like this:
100 cells
and each cell is about 20km wide, while a 500-cells map would look like this:
500 cells
where each cell is 4km wide and you really need to zoom in to see poseidon objects: 500 cells zoomed
In general the more cells the slower the model runs because of the effort in pathfinding for boats and fish movement. But having it scalable is important for general testing of the model.

Biology

As with most other POSEIDON projects, it pays to cannibalize on old code when we can, especially regarding biology. Here I take most of the infrastructure developed for our West Coast validation model, fill it with silly numbers and place it in Indonesia.
Assume fish is of one species, “Red Fish” (we can turn it into snapper later when parametrizing), and we split it into 3 age bins:

Bin Weight Maturity Movement
0 100 g 0 no
1 1kg 0.5 yes
2 2kg 1 yes

Every year 20% of each age bin grows older (or die if it’s in the last bin). Spawning is Beverton-Holt, assuming 40M virgin recruits, steepness of 0.6 and \(\Phi\) of 14.24 (these are just the numbers for Sablefish in California). Red fish only live in depths of between 800 and 1500 meters.

Mature red fish, i.e. bin 1 or 2, also moves. It does so in a very simple diffusing rate where, for any two contiguous cells if there is a differential in abundance for that bin, fish move from the cell with most fish to the one with least: \[ \text{Movement}_{ij} = \max(0, m \left( \text{Abundance}_i - \text{Abundance}_j \right)) \] Where \(m\) is the diffusing rate, capped at 0.001 for this simulation.

I am not supposing red fish even remotely resembles snappers, my objective here is to test all the individual components (movement, spawning, allocation, aging) to get them ready for more realistic work.
A secondary objective however is to generate a simple scenario where red fish disperses in a way that makes policy interesting.

Let’s assume that for whatever reason at year 0, you just have 40M fish at age 0 in a few cells. Further assume that any new recruit spawned in the next years will start its life at these initial “nursery” cells. Mature fish will start moving out of the nursery cells expanding into neighboring areas of the right depth.

In the next video I just plot biomass of the red fish in each cell over the years. After the first year (where all the biomass is in the nursery area) you can start seeing biomass moving out. This creates a nice setup where most of the biomass is still in the nursery, but if fishers were somehow incentivated to fish outside of the nursery they would end up catching only fish of age 1 or 2.

Fishery

Imagine now that there are 50 boats in Kupang. They are small, they carry only 100kg of fish in the hold at max, they have a catchability of 0.0001 for all age bins and gas is expensive so that they prefer to fish near port rather than far from it. Imagine that fish is paid 10$ a kg regardless of the bin caught. Then obviously fishers go somewhere that is close and abundant (that is, they go to fish in the nursery).

Now imagine that we impose a fine of 1000$/kg on catching age 0 fish and put observers in every boat. At that point agents will react by fishing on the “nursery line”, catching only age 1-2 fish as it moves out of the nursery.

If you impose a fine on age 0 and age 1, fishers are in a bind because age 1 and age 2 fish both move and live in the same areas. However there is a lot of age 1 fish moving out of nurseries so that the proportion age 2 to age 1 (that is the proportion of allowed vs banned) is higher the further from the nursery one goes fishing. Fishers in POSEIDON figure this out and you see them avoiding the nursery line by fishing at the edge of the fish habitat.

Policy

Imagine we plan on reducing the number of age 0 catches by using market incentives. We can either tax catches of age 0 fish or subsidise catches for more mature landings.

Observers on board

It is worth noting that with no market intervention only about 25% of the landed weight is age 0. Imagine we can impose any tax up to a fine of 1000% to landed weight. We run the simulation once for each 0.1$ step and these are the results:

I am showing an average out of 3 runs so that the real effect is noisy but you can clearly tell that fines have an effect on how much fishers target young biomass. Unfortunately it looks like fines have to be quite steep to have any effect (given our made up parameters, of course).

What about subsidies?

Subsidies also have an effect but it is more muted: juveniles always make up at least 10% of all the mass landed.

The slope of these curve depends fundamentally on the parameters used so I don’t think we should learn anything out of this exercise except how to carry it out.

Discarding

Now of course I made a silly assumption there by assuming a perfectly enforced discard ban. What if agents can decide endogenously whether to discard unsellable fish or not? Easiest way to do it in POSEIDON is simply to assume again explore-exploit-imitate where every two months people who discard look at people who don’t discard (and viceversa) to see if they are making more money.

Let’s take again the example where we impose a fine. If we just look at landing rate we say: look what a great success our fines have been, people stopped landing young fish with just a tiny little tax!

However if we look at discard rate (defined here as % of fishers whose strategy involves throwing back at sea dead juveniles): Of course it’s extremely high as soon as any tax is introduced.

What if we went the other route and offered subsidies?

Again we get a much stronger reaction to the policy if we look at landings and the difference is once again due to people discarding more.